Approaching Object Detector Random Behavior Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting objects approaching a vehicle using optical flow from camera images often result in false detections due to high-speed vehicle motion, edge features like white lines, and noise, leading to inaccurate identification of approaching objects.
Innovation Solution
An approaching-object detector that extracts feature points from chronologically captured images, detects their behavior, and determines if it's random, excluding feature points with random behavior to improve detection accuracy by only considering points with regular behavior for approaching object determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If optical flow is used to detect moving objects from camera images, then object detection capability is provided, but false detections increase due to high-speed vehicle motion, edge features, and noise
Solution Approach 1:
The patent changes the parameter of feature point selection by introducing randomness evaluation. Instead of using all detected feature points, the system evaluates whether each feature point exhibits random behavior and excludes those that do. This parameter change in the detection process reduces false detections caused by noise and edge features while maintaining true object detection capability.
2Measurement precision
If all feature points are used for object detection, then detection coverage is maximized, but detection accuracy decreases due to inclusion of random behavior points
Solution Approach 1:
The patent applies local quality by treating different feature points differently based on their individual behavior characteristics. Each feature point is evaluated for randomness, and those exhibiting random behavior are excluded while those with regular behavior are retained. This localized quality assessment improves overall detection accuracy without unnecessarily reducing the total number of useful feature points.
3Reliability
If feature points with random behavior are included in detection, then computational complexity is reduced, but detection reliability decreases
Solution Approach 1:
The patent implements self-service by enabling feature points to evaluate themselves for random behavior. Each feature point undergoes an randomness evaluation process where its own behavior patterns are analyzed. This self-assessment mechanism automatically filters out unreliable feature points without requiring complex external validation systems, thereby improving detection reliability with manageable computational complexity.
Data Source
AI summary
An approaching-object detector for detecting an object approaching an own vehicle includes: a memory; and a processor configured to perform a process, the process including extracting a plurality of corresponding feature points from chronologically captured images, which are obtained by capturing the object using an imaging device provided for the own device, detecting a behavior among the captured images in regard to each of the plurality of feature points, determining whether or not the behavior is random in regard to each of the plurality of feature points, and determining whether or not the object is approaching the own vehicle based on a behavior of a feature point whose behavior is determined to be not random among the plurality of feature points, and outputting a result of the determination.